ax_action_recognition

ax_action_recognition

Input

Video

(Video is generated from dataset in https://github.com/felixchenfy/Realtime-Action-Recognition)

Input shape : (1, C, T, V, M) = (1, 2, 15, 18, 1)

C: channel number
T: frame number
V: keypoint number
M: person ID

Output

Category

CATEGORY = (
'stand', 'walk', 'run', 'jump', 'sit', 'squat', 'kick', 'punch', 'wave'
)

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

By adding the --video option, It can be run as a real-time mode that infers frame by frame of the video. If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 ax_action_recognition.py --video VIDEO_PATH -a lw_human_pose
$ python3 ax_action_recognition.py --video VIDEO_PATH -a pose_resnet

Reference

• Realtime-Action-Recognition
• ST-GCN

The architecture of ax_action_recognition model is simplified from ST-GCN, and trained with Realtime-Action-Recognition dataset.

Framework

Pytorch

Model Format

ONNX opset=11

Netron

ax_action_recognition.onnx.prototxt